TOPPAN Holdings, Aisin, Asahi Kasei, SoftBank, AGC and DIC announced on October 28, 2026 that they will conduct joint research on materials development using quantum computers and AI. The project will run for three years, from fiscal 2026 to fiscal 2028, and will target next-generation batteries, optically functional materials, superconducting materials and more. The companies plan to combine quantum computation, conventional computational methods and AI to predict material properties, then compare those predictions with measurements taken from materials that are actually synthesized. Research on using quantum computers in materials development is advancing, but how much this project will improve prediction accuracy and development efficiency remains to be tested. By combining computation with experiments, the six companies aim to establish a development method that can be applied across a wide range of materials fields.
Combining quantum computers and AI to predict material properties
In developing new materials, being able to predict properties before synthesis makes it possible to narrow down promising candidates and make experiments more efficient. This joint research will test how effective it is to combine quantum computation, conventional computational methods and AI for this kind of property prediction. As background, the six companies cite the difficulty of securing the high-quality data needed to train AI, and the fact that conventional computational methods require large amounts of computing resources when scaling up accuracy or size. Quantum computers may be able to simulate the behavior of molecules and electrons more efficiently than before, and the project will explore whether this characteristic can be used to generate data and make predictions for materials development.
However, today's quantum computers face constraints such as the number of usable qubits and errors that occur during computation, making it difficult for them to handle large-scale materials simulations on their own. The study therefore does not rely on quantum computers alone, but combines them with supercomputers and AI. SoftBank is responsible for a computing platform that links quantum computers and supercomputers, and will verify a mechanism for using the generated data in AI. Quantum computation has not solved the problems of data scarcity or computational cost; the project is at the stage of investigating what improvements can be gained by combining it with existing computing technologies.
Six companies share roles, testing everything from batteries to superconducting materials
On July 17, 2026, the New Energy and Industrial Technology Development Organization (NEDO) announced the implementation structure for a large-scale demonstration aimed at creating use cases under its Post-5G project. On the first page of the appendix, the six companies taking part in this joint research are listed as implementers of the same research theme. The October 28 announcement revealed the signing of an agreement following that selection, along with the specific research content and each company's role.
According to the joint announcement by the six companies, SoftBank is in charge of the platform linking quantum computers and supercomputers, while the other companies will verify applications to materials in their respective areas of expertise. The research targets and roles of each company are as follows.
| Company | Target materials | Planned research |
|---|---|---|
| TOPPAN Holdings | Additive materials | Combine quantum and conventional algorithms; examine integration with AI and feedback of experimental results |
| Aisin | Superconducting materials | Apply quantum computation and compare it with existing computational methods |
| Asahi Kasei | Organic compounds such as electrolytes, polymers, and inorganic materials such as electrodes | Verify the applicability of quantum computation to various materials |
| SoftBank | Next-generation batteries and others | Generate data by linking quantum computers and supercomputers, and demonstrate its use by AI |
| AGC | Inorganic materials such as optical glass, and organic materials such as electronic materials | Combine conventional computational methods, AI and quantum computation to build guidelines for materials design |
| DIC | Colorants, semiconductor materials | Examine research methods for predicting properties before materials are synthesized |
The research targets range widely, from battery materials to superconducting materials, optical glass, colorants and semiconductor materials. However, the properties to be predicted and the accuracy required differ by material, so the same method will not necessarily work for all of them. For example, Aisin will apply quantum computation to superconducting materials and compare it with conventional methods, while DIC will examine methods for predicting the properties of colorants and semiconductor materials before synthesis. Verification across these different fields will be important for clarifying what kinds of problems quantum computation is suited to. The inclusion of multiple materials as research targets does not in itself mean that a general-purpose development platform capable of handling every material has been completed.
Checking computational predictions against actual material synthesis
In addition to integrating quantum computation and AI, the joint research plans to verify applicability to various materials and to compare the results of material synthesis and evaluation with simulation results. NEDO's large-scale demonstration is also positioned as an effort that emphasizes demonstration in real-world environments, not just verification on computers. In the research, material properties will first be predicted using computation and AI, and the materials will then be actually synthesized and the measured values compared with the predictions. This comparison is intended to evaluate how accurately computational predictions capture the properties of real materials.
In materials development, a candidate judged promising in simulation does not always show the expected properties when actually synthesized. Comparing computational and experimental results is therefore important for judging how effective a prediction method is. If predicted values are close to measured values, they may help in selecting candidates to take to experiments, but how much accuracy can be achieved depends on future verification. Whether incorporating quantum computation yields better results than conventional methods has also not been concluded at this point.
The research will also examine a mechanism for reflecting results obtained from experiments in subsequent calculations and predictions. TOPPAN Holdings includes a method for feeding experimental results back into computation among its research items, and the six companies aim to build a continuously improving platform that integrates quantum computation, AI and experimental data. The idea is not to stop at predicting material properties by computation, but to apply knowledge gained from experiments to the next prediction and steadily raise the precision of materials development. However, the announcement did not reveal specific terms, such as how data will be shared among the companies or how intellectual property from the research results will be handled.
Judging the practical value of quantum computation requires comparison with conventional methods
What matters in this joint research is not the fact that quantum computers were used, but what improvements they bring to materials development. For example, comparing a method that incorporates quantum computation with a conventional computational method, using the same material and the same level of prediction accuracy as conditions, would offer a clue to the effect of adding quantum computation. However, this is a comparison approach considered necessary when verifying research results; the six companies have not announced that they have decided on specific comparison conditions.
Even if computational predictions agree with experimental results, that alone does not prove the effectiveness of quantum computation. If the same prediction accuracy can also be achieved with conventional methods, the additional advantages of using quantum computation need to be shown. Evaluation involves not only prediction error but also computation time, required computing resources, and effectiveness in narrowing down material candidates for experiments. Even if part of the computation becomes faster, the time and cost of the entire development process, from material selection through synthesis to evaluation, will not necessarily fall.
What has been made public so far consists of the six companies' joint press release and NEDO's project materials, not a research report presenting specific experimental results. The number of samples to be evaluated, the compositions of the materials, the quantum computer models to be used and the measurement conditions have not been disclosed. No quantitative data have been published on how much computation time or prediction accuracy is expected to improve. The published materials also give no information on peer-reviewed papers or preprints corresponding to this research, nor on replication results from independent research institutions, so performance and reproducibility cannot be judged on the basis of measured data.
In the joint research running from fiscal 2026 to fiscal 2028, in addition to comparing methods that include quantum computation with conventional ones, how closely predicted results match actual material properties will be important. If these verification results are published with specific figures and experimental conditions, it will become possible to evaluate more objectively how much the combination of quantum computers and AI contributes to more efficient materials development. The initiative can be seen as an attempt to verify the practical value of quantum computation not only through computational performance but through actual results in materials development.